DOCSLIB.ORG

Simulation Experiments: Better Data, Not Just Big Data

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Simulation Experiments: Better Data, Not Just Big Data Proceedings of the 2015 Winter Simulation Conference L. Yilmaz, W. K V. Chan, I. Moon, T. M. K. Roeder, C. Macal, and M. D. Rossetti, eds. SIMULATION EXPERIMENTS: BETTER DATA, NOT JUST BIG DATA Susan M. Sanchez Naval Postgraduate School Operations Research Department 1411 Cunningham Rd. Monterey, CA 93943-5219, USA ABSTRACT Data mining tools have been around for several decades, but the term “big data” has only recently captured widespread attention. Numerous success stories have been promulgated as organizations have sifted through massive volumes of data to find interesting patterns that are, in turn, transformed into actionable information. Yet a key drawback to the big data paradigm is that it relies on observational data—limiting the types of insights that can be gained. The simulation world is different. A “data farming” metaphor captures the notion of purposeful data generation from simulation models. Large-scale designed experiments let us grow the simulation output efficiently and effectively. We can explore massive input spaces, uncover interesting features of complex simulation response surfaces, and explicitly identify cause-and-effect relationships. With this new mindset, we can achieve quantum leaps in the breadth, depth, and timeliness of the insights yielded by simulation models. 1 INTRODUCTION There is no universally accepted definition for the term “big data.” Many have argued that we have always had big data, whenever we had more available than we could analyze in a timely fashion with existing tools. Running a single multiple regression in the 1940’s was very difficult, since matrix multiplication and inversion for even a small problem (say, a handful of independent variables and a few hundred observations) is a non-trivial task when it must be done by hand. From that perspective, big data can be viewed as any data set that pushes against the limits of currently available technology. What can be done with big data? Most people immediately think of data mining, a term that is ubiquitous in the literature. The concept of data farming is less well-known, but has been used in the defense community over the past decade. At the Simulation Experiments & Efficient Designs (SEED) Center for Data Farming at the Naval Postgraduate School, we describe the differences between these metaphors as follows (Sanchez 2010, Sanchez 2014). Miners seek valuable nuggets of ore buried in the earth, but have no control over what’s out there or how hard it is to extract the nuggets from their surroundings. As they take samples from the earth, they gather more information about the underlying geology. Similarly, data miners seek to uncover valuable nuggets of information buried within massive amounts of data. They may look at “all” the “big data” available at a particular point in time, but they typically have no control over what data are out there, nor how hard it will be to separate the useful information from the rest of the data. Data mining techniques use statistical and graphical measures to try to identify interesting correlations or clusters in the data set. Farmers cultivate the land to maximize their yield. They manipulate the environment to their advantage, by using irrigation, pest control, crop rotation, fertilizer, and more. Small-scale designed experiments can help them determine whether these treatments are effective. Similarly, data farmers manipulate simulation models to advantage—but using large-scale designed experimentation. This allows them to learn more about the simulation model’s behavior in a structured way. In this fashion, they “grow” data from their 978-1-4673-9743-8/15/$31.00 ©2015 IEEE 800 Sanchez models, but in a manner that facilitates identifying the useful information. For large-scale simulation experiments, this often results in big data sets. Yet, although the data sets are big, they are far smaller than what would be needed to gain insights if the results were observational (i.e., obtained using ad hoc or randomly generated combinations of factor settings). The data sets are also better, in the sense they let us identify root cause-and-effect relationships between the simulation model input factors and the simulation output. 2 REVISITING CAUSATION AND CORRELATION The simplest way of establishing cause-and-effect is via an experiment. Let’s suppose that we’re looking at a stochastic system, so our observation of Y comes with some error. An experiment involving a single factor X involves changing X and observing whether or not there is a change in the response Y. We’re trying to uncover the ground truth about the relationship between X and Y over a range of interest, as shown in (1). Y = g(X)+ ε(X) (1) After collecting data, we will end up estimating this via some metamodels, as shown in (2). The subscripts ong ˆ and εˆ show these models were estimated from data obtained using design D. Yˆ = gˆD(X)+ εˆD(X) (2) Three important concepts in the design of experiments (DOE) are: control, randomization, and repli- cation. For real-world experiments, you exercise control over the situation by deciding which values of X are of interest. You also decide how to control for everything else that isn’t of interest, perhaps by holding it constant or using a control group for comparison purposes. Randomization is used to guard against hidden or uncontrollable sources of bias. For example, if you are measuring the miles per gallon of different vehicles by using a single driver, randomizing the order in which they are driven will remove any systematic bias due to fatigue. With replication you collect multiple observations to assess the magnitude of the variability associated with Y, so you can construct confidence intervals or conduct significance tests. In simulation experiments, we use these concepts in different ways. In the simulation world, the analyst has total control. Potential factors in simulation experiments include the input parameters or distributional parameters of a simulation model, whether they are controllable in the real world or not. The analyst also has control over the random number seeds and streams. This means that, unlike physical experiments where uncontrollable noise occurs, the results from a simulation experiment are perfectly repeatable, and randomization is not needed to guard against hidden or uncontrollable sources of bias. Replication means you get multiple experimental units (runs or batches) to get a sense of the magnitude of the variability associated with Y. There are certainly times when uncovering correlation is very useful. As Hogan (2014) points out, “Epidemiological studies demonstrated more than a half century ago a strong correlation between smoking and cancer. We still don’t understand exactly how smoking causes cancer. The discovery of the correlation nonetheless led to anti-smoking campaigns, which have arguably done more to reduce cancer rates over the past few decades than all our advances in testing and treatment.” Yet there are also many more times when identifying correlation is not nearly good enough. It’s well-known that trolling through mountains of data can yield spurious correlations: an excellent discussion of how this affects published research results appears in The Economist (2013). If big data sets reveal a correlation between two variables X and Y, the ground truth can be one of four basic situations: (i) changes in X cause changes in Y; (ii) changes in Y cause changes in X; (iii) changes in X and Y are both the result of changes in other, potentially unknown or unobservable factors; or (iv) this is a spurious correlation. One drawback of big observational data is that we have no real way of testing these results without moving to a different environment. Common sense can help eliminate some spurious correlations (NPR 2014, Vigen 2014), but one of the tenets of big data is that it is not repeatable. Correlations that appear strong at one point in time and weak at another might mean that the underlying system has changed. 801 Sanchez In their recent book, Mayer-Schonberger¨ and Cukier (2013) claim that “the need for sampling is an artifact of a period of information scarcity” and that big data “leads society to abandon its time-honored preference for causality, and in many instances tap the benefits of correlation.” But if the goal of analysis is to yield better outcomes via controlling or influencing the inputs, what could be more important than causality? 3 CHARACTERIZING BIGNESS Why are so many either extolling or lamenting big data’s focus on correlation rather than causation? I suggest it is because their view of big data is observational. This is not the case for those using simulation. Using the data farming metaphor, we grow the data for analysis, rather than mine existing data. However, just as a different mindset is needed for dealing with big observational data, so a different mindset is needed for generating and dealing with big simulation data. 3.1 The 3 (or more) V’s of Big Data Big data is typically characterized by 3 V’s — volume, velocity, and variety (Laney 2001). These have all increased at an astonishing rate during the last decade, with much of the data being generated, captured from, or stored on the internet. Volume refers to the amount of data, and this is enormous. As of June 2014, IBM states that “90% of the data in the world today has been created in the last two years alone” (IBM 2015), but that is already out of date. Some have advocated a look at more V’s (including veracity, validity, and volatility, viability, value, and victory), but others argue that these are derived from context-specific analytics while the original 3 V’s capture the “intrinsic, definitional big data properties essence of big data” (Grimes 2013).
Load more

Recommended publications
  • Micro-Services for Facilitating Data Farming in Nato
    Proceedings of the 2019 Winter Simulation Conference N. Mustafee, K.-H.G. Bae, S. Lazarova-Molnar, M. Rabe, C. Szabo, P. Haas, and Y.-J. Son, eds. DATA FARMING SERVICES: MICRO-SERVICES FOR FACILITATING DATA FARMING IN NATO Daniel Huber Gary Horne Fraunhofer IAIS MCR Global Schloss Birlinghoven, Avenue du Bourget 42 53757 Sankt Augustin, GERMANY 1130 Brussels, BELGIUM Daniel Kallfass Jan Hodicky Airbus Defence and Space GmbH Aviation Technology Department Claude-Dornier-Straße University of Defence 88090 Immenstaad, GERMANY Kounicova 65 Brno, 60200, CZECH REPUBLIC Nico de Reus TNO - Netherlands Organisation for Applied Scientific Research Oude Waalsdorperweg 63 The Hague, 2509 JG, THE NETHERLANDS ABSTRACT In this paper the concept and architecture of Data Farming Services (DFS) are presented. DFS is developed and implemented in NATO MSG-155 by a multi-national team. MSG-155 and DFS are based on the data farming basics codified in MSG-088 and the actionable decision support capability developed in MSG-124. The developed services are designed as a mesh of microservices as well as an integrated toolset and shall facilitate the use of data farming in NATO. This paper gives a brief description of the functionality of the services and the use of them in a web application. It also presents several use cases with current military relevance, which were developed to verify and validate the implementation and define requirements. The current prototype of DFS was and will be tested at the 2018 and 2019 NATO CWIX exercises, already proving successful in terms of deployability and usability. 1 INTRODUCTION 1.1 Data Farming in NATO Data Farming is a quantified approach that examines questions in large possibility spaces using modeling and simulation.
    [Show full text]
  • Data Farming Services in Support of Military Decision Making
    Data Farming Services in Support of Military Decision Making Nico de Reus TNO - The Netherlands Organisation for Applied Scientific Research Oude Waalsdorperweg 63, 2597 AK The Hague THE NETHERLANDS [email protected] Maj. Ab de Vos Ministry of Defence JIVC/KIXS Herculeslaan 1 3584 AB Utrecht THE NETHERLANDS [email protected] LtCdr Bernt Åkesson Finnish Defence Research Agency P.O. Box 10, FI-11311, Riihimäki FINLAND [email protected] Gary Horne MCR 2010 Corporate Ridge, McLean, Virginia 22102 USA [email protected] Maj. Tobias Kuhn M&S Centre of Excellence Piazza Renato Villoresi 1, 00143 Rome ITALY [email protected] Viktor Stritof, CZU/ERS CZU/ERS Ljubljanska cesta 37, 6230 Postojna SLOVENIA [email protected] LtCol Stephan Seichter Bundeswehr Office for Defence Planning Einsteinstrasse 20, 82024 Taufkirchen GERMANY [email protected] Alexander Zimmermann, Kai Pervölz Fraunhofer IAIS Schloss Birlinghoven, 53757 Sankt Augustin GERMANY [email protected], [email protected] STO-MP-IST-160 PP-5 - 1 Data Farming Services in Support of Military Decision Making ABSTRACT Data Farming is a quantified approach that examines questions in large possibility spaces using modelling and simulation. By harvesting simulation data from many runs set up in a cogent manner, the data farming process evaluates huge amounts of simulation data to draw insights to support military decision making. Data Farming has been codified and methods for actionable decision support have been developed in the MSG-088 and MSG-124 Task Groups. Currently, the new MSG-155 Task Group has started with plans to make Data Farming accessible and usable by NATO and its partners through the development of Data Farming Services.
    [Show full text]
  • Data Farming: Better Data, Not Just Big Data
    Proceedings of the 2018 Winter Simulation Conference M. Rabe, A. A. Juan, N. Mustafee, A. Skoogh, S. Jain, and B. Johansson, eds. DATA FARMING: BETTER DATA, NOT JUST BIG DATA Susan M. Sanchez Naval Postgraduate School Operations Research Department 1411 Cunningham Rd. Monterey, CA 93943-5219, USA ABSTRACT Data mining tools have been around for several decades, but the term “big data” has only recently captured widespread attention. Numerous success stories have been promulgated as organizations have sifted through massive volumes of data to find interesting patterns that are, in turn, transformed into actionable information. Yet a key drawback to the big data paradigm is that it relies on observational data, limiting the types of insights that can be gained. The simulation world is different. A “data farming” metaphor captures the notion of purposeful data generation from simulation models. Large-scale experiments let us grow the simulation output efficiently and effectively. We can use modern statistical and visual analytic methods to explore massive input spaces, uncover interesting features of complex simulation response surfaces, and explicitly identify cause-and-effect relationships. With this new mindset, we can achieve tremendous leaps in the breadth, depth, and timeliness of the insights yielded by simulation. 1 INTRODUCTION What can be done with big data? Most people immediately think of data mining, a term that is ubiquitous in the literature (see, e.g., James et al. 2013 or Witten et al. 2017 for overviews). The concept of data farming is less well known, but has been used in the defense community over the past two decades (Brandstein and Horne 1998).
    [Show full text]
  • Big Data in Smart Farming – a Review
    Agricultural Systems 153 (2017) 69–80 Contents lists available at ScienceDirect Agricultural Systems journal homepage: www.elsevier.com/locate/agsy Review Big Data in Smart Farming – A review Sjaak Wolfert a,b,⁎,LanGea, Cor Verdouw a,b, Marc-Jeroen Bogaardt a a Wageningen University and Research, The Netherlands b Information Technology Group, Wageningen University, The Netherlands article info abstract Article history: Smart Farming is a development that emphasizes the use of information and communication technology in the Received 2 August 2016 cyber-physical farm management cycle. New technologies such as the Internet of Things and Cloud Computing Received in revised form 31 January 2017 are expected to leverage this development and introduce more robots and artificial intelligence in farming. Accepted 31 January 2017 This is encompassed by the phenomenon of Big Data, massive volumes of data with a wide variety that can be Available online xxxx captured, analysed and used for decision-making. This review aims to gain insight into the state-of-the-art of Keywords: Big Data applications in Smart Farming and identify the related socio-economic challenges to be addressed. Fol- Agriculture lowing a structured approach, a conceptual framework for analysis was developed that can also be used for future Data studies on this topic. The review shows that the scope of Big Data applications in Smart Farming goes beyond Information and communication technology primary production; it is influencing the entire food supply chain. Big data are being used to provide predictive Data infrastructure insights in farming operations, drive real-time operational decisions, and redesign business processes for Governance game-changing business models.
    [Show full text]
  • Types of Data & Terms Related with Data
    Modeling and Simulation Body of Knowledge (M&SBOK) - Index updated and © by: Dr. Tuncer Ören - 2010-12-17 Terminology: Types of Data & Terms Related with Data (based on the first version of the Multi-lingual Modeling and Simulation Dictionary) accessible data actual data adjusted data agent-based data mining agent-based distributed data mining ALSP protocol data ambiguous data analytical data compression assessment of data assessment of real-system data assessment of simulated data auditable data authentication data authoritative data source battlespace database behavior data base big endian data format bivariate data calibrated data calibration data certification data certified data coarse data common database complex data conceptual data model consistent data correct data cultural features data current data data data collection data currency data interchange data standardization data steward data synchronization data acceptability data accessibility data accuracy data acquisition data administration data administrator data aggregation data analysis data animation data appropriateness data architecture data association data attribute data authentication data availability data- based data center data certification data characteristic data collection assessment data collectıon phase data collector data completeness data compression data consistency data consumed by M&S data consumer data correctness data cost data coupling data customer data dependency analysis data dictionary data dictionary system data- directed data discrepancy data
    [Show full text]
  • Data Mining What Is Data Mining? Data Mining “Architecture
    What is Data Mining? Data Mining • Domain understanding • Data selection Andrew Kusiak Intelligent Systems Laboratory • Data cleaning, e.g., data duplication, 2139 Seamans Center missing data The Un ivers ity o f Iowa • Preprocessing, e.g., integration of different Iowa City, IA 52242 - 1527 [email protected] files http://www.icaen.uiowa.edu/~ankusiak • Pattern (knowledge) discovery Tel. 319-335 5934 Fax. 319-335 5669 • Interpretation (e.g.,visualization) • Reporting Data Mining “Architecture” Illustrative Applications • Prediction of equipment faults • Determining a stock level • Process control • Fraud detection • Genetics • Disease staging and diagnosis • Decision making Pharmaceutical Industry Pharmaceutical Industry • Selection of “Patient suitable” medication – Adverse drug effects minimized An individual object (e.g., product, – Drug effectiveness maximized patid)iiient, drug) orientation – New markets for “seemingly ineffective ” drugs vs • “Medication bundle” – Life-time treatments A population of objects (products, patients, drugs) orientation • Design and virtual testing of new drugs 1 What is Knowledge Discovery? Data Mining is Not • Data warehousing Data • SQL / Ad hoc queries / reporting Set • SftSoftware agen ts • Online Analytical Processing (OLAP) • Data visualization E.g., Excel, Access, Data Warehouse Learning Systems (1/2) Learning Systems (2/2) • Classical statistical methods (e.g., discriminant analysis) • Association rule algorithms • Modern statistical techniques • Text mining algorithms (e.g., k-nearest
    [Show full text]
  • UI for Radiation Therapy Cohort Selection Seminar Presentation
    Project 8: UI for Radiation Therapy Cohort Selection Seminar Presentation Members: Domonique Carbajal, Keefer Chern Mentors: Todd McNutt,Pranav Lakshminarayanan TM CIS II 601.456 Spring 2019 Project Objective Develop a User Interface that will allow user the ability to: • Select a patient cohort based upon any number of variables (SQL) • Perform statistical analysis on the extracted data • Display the data in an easily comprehensible way (C# and JavaScript) • Load and save parameters in a database query call CIS II 601.456 Spring 2019 Paper Selection The Big Data Effort in Radiation Oncology: Addresses conceptualization of handling data specifically for Data Mining or Data Farming? radiation oncology and poses Mayo, Charles S., et al. “The Big Data Effort in Radiation Oncology: Data Mining or Data Farming?” Advances in Radiation Oncology, vol. 1, no. 4, 2016, warnings for handling errors in the pp. 260–271., doi:10.1016/j.adro.2016.10.001. database Complete Author List: CIS II 601.456 Spring 2019 The Big Data Effort in Radiation Oncology: Data Mining or Data Farming? Purpose Describe data related issues in ROIS, impart vision for solutions and key data elements that need to be addressed for fully utilizing available information . Introduce metaphor of “data farming” and necessary distinctions from data mining CIS II 601.456 Spring 2019 Terminology Used and Defined PQI - Practical Quality Improvement ROIS - Radiation Oncology Information System ETL - Extract, Transform, Load Data Mining - (As defined by paper) Data aggregation and analysis efforts, “mining” creates expectations data elements needed already exist in electronic system Assumes data allows for accurate linkage to patients, identification of relationships among data elements, and extraction of reliable values M-ROAR - University of Michigan instance of a Radiation Oncology Analytics Resource CIS II 601.456 Spring 2019 Mayo, Charles S., et al.
    [Show full text]
  • Establishing a Data Farm to Harvest Quality Information
    Paper Number LPS-3C Establishing a Data Farm to Harvest Quality Information Harold W Thomas (Hal) Air Products and Chemicals, Inc. Mike Moosemiller Det Norske Veritas Prepared For 35th Annual Loss Prevention Symposium AIChE 2001 Spring National Meeting Houston, TX April 22-26 2001 November 2000 Unpublished AIChE shall not be responsible for statements or opinions contained in papers or printed in its publications. Copyright © 2001 AIChE Page 1 or 19 Abstract To obtain useful reliability information from maintenance and inspection data once it has been collected requires effort. In the past, this has been referred to as data “mining”, as if the data can be extracted in its desired form if only it can be found. In contrast, this paper proposes data “farming”, and describes the seeds that are necessary to harvest the best possible crop of reliability information. The CCPS equipment reliability database project provides valuable lessons on how to “farm” rather than merely “mine” data. The CCPS work processes for establishing failure modes, populations to track, event data to collect, and implementation are all reviewed. Attention is given to knowing up front the data objectives and the quality of information desired. Also, the treatment of equipment surveillance periods turns out to be a critical variable for data quantity and quality; reasons for this and approaches to take are discussed. It will be seen that the quality and continuity of information that can be derived is much greater when the data sources can be “farmed” rather than “mined”.
    [Show full text]
  • Data Farming and Quantitative Analysis of Cyber Defense Technologies and Measures
    MODSIM World 2016 Data Farming and Quantitative Analysis of Cyber Defense Technologies and Measures Gary Edward Horne Santiago Balestrini Robinson Blue Canopy Group Georgia Tech Research Institute Reston, Virginia Atlanta, Georgia [email protected] [email protected] ABSTRACT Data Farming is a quantified approach that examines questions in large possibility spaces using modeling and simulation. It evaluates whole landscapes of outcomes to draw insights from outcome distributions and outliers. The Data Farming Support to NATO task group has codified the data farming methodology and a follow-on task group is now applying data farming to important NATO question areas. The development of data mining and data visualization techniques also continues to help us understand the huge amount of simulation data resulting from data farming. This paper outlines the documented data farming techniques, illustrates data farming in the context of quantitative analysis of cyber defense technologies and measures, and describes the links to predictive analyses made possible by advancing the connection of data farming to data mining and data visualization. The paper describes a prototype simulation using an agent-based model (NetLogo) that the NATO task group team has developed with feedback from subject-matter experts. The paper also describes the data farming the team has performed on key model parameters to begin to get insight into cyber security “what-if” questions. ABOUT THE AUTHORS Gary Edward Horne is a Research Analyst at Blue Canopy Group with a doctorate in Operations Research from The George Washington University. During his career in defense analysis, he has led data farming efforts examining questions in areas such as humanitarian assistance, convoy protection, and anti-terrorist response.
    [Show full text]
  • Visualization and Interaction for Knowledge Discovery in Simulation Data
    Proceedings of the 53rd Hawaii International Conference on System Sciences | 2020 Visualization and Interaction for Knowledge Discovery in Simulation Data Niclas Feldkamp Sören Bergmann Thomas Schulze Steffen Strassburger Otto-von-Guericke-Universität Magdeburg Technische Universität Ilmenau [email protected] {niclas.feldkamp, soeren.bergmann, steffen.strassburger}@tu-ilmenau.de Abstract influential on the project scope [21]. Kleijnen describes this as the trial-and-error approach for finding good Discrete-event simulation is an established and solutions and that simulation users should spent more popular technology for investigating the dynamic time analyzing their models or model output, behavior of complex manufacturing and logistics respectively. Besides, the development of simulation systems. Besides traditional simulation studies that models is a very time-consuming and costly process, so focus on single model aspects, data farming describes that the potential of the model should be leveraged as an approach for using the simulation model as a data much as possible [21]. generator for broad scale experimentation with a With the widespread availability of Big Data broader coverage of the system behavior. On top of infrastructures and ease of access to computing power, that we developed a process called knowledge there are new possibilities for leveraging the potential discovery in simulation data that enhances the data of simulation models. In this regard, we developed a farming concept by using data mining methods for the concept called Knowledge Discovery in Simulation data analysis. In order to uncover patterns and causal Data that allows to use a combination of data mining relationships in the model, a visually guided analysis and visual analysis in order to find hidden and then enables an exploratory data analysis.
    [Show full text]
  • Big Data in Smart Farming – a Review
    Agricultural Systems 153 (2017) 69–80 Contents lists available at ScienceDirect Agricultural Systems journal homepage: www.elsevier.com/locate/agsy Review Big Data in Smart Farming – A review Sjaak Wolfert a,b,⁎,LanGea, Cor Verdouw a,b, Marc-Jeroen Bogaardt a a Wageningen University and Research, The Netherlands b Information Technology Group, Wageningen University, The Netherlands article info abstract Article history: Smart Farming is a development that emphasizes the use of information and communication technology in the Received 2 August 2016 cyber-physical farm management cycle. New technologies such as the Internet of Things and Cloud Computing Received in revised form 31 January 2017 are expected to leverage this development and introduce more robots and artificial intelligence in farming. Accepted 31 January 2017 This is encompassed by the phenomenon of Big Data, massive volumes of data with a wide variety that can be Available online xxxx captured, analysed and used for decision-making. This review aims to gain insight into the state-of-the-art of Keywords: Big Data applications in Smart Farming and identify the related socio-economic challenges to be addressed. Fol- Agriculture lowing a structured approach, a conceptual framework for analysis was developed that can also be used for future Data studies on this topic. The review shows that the scope of Big Data applications in Smart Farming goes beyond Information and communication technology primary production; it is influencing the entire food supply chain. Big data are being used to provide predictive Data infrastructure insights in farming operations, drive real-time operational decisions, and redesign business processes for Governance game-changing business models.
    [Show full text]
  • Data Farming: Methods for the Present, Opportunities for the Future
    DATA FARMING: METHODS FOR THE PRESENT, OPPORTUNITIES FOR THE FUTURE Susan M. Sanchez SEED Center for Data Farming Operations Research Department Naval Postgraduate School ISIM 2017 Research Workshop Durham, U.K. Department of Defense Distribution Statement: Approved for public release; distribution is unlimited Data Mining vs. Data Farming 2 Data Mining vs. Data Farming • Miners seek valuable buried nuggets 2 Data Mining vs. Data Farming • Miners seek valuable buried nuggets - Miners have no control over what’s there or how hard it is to separate it out 2 Data Mining vs. Data Farming • Miners seek valuable buried nuggets - Miners have no control over what’s there or how hard it is to separate it out - Data Mining seeks valuable information buried within massive amounts of data 2 Data Mining vs. Data Farming • Miners seek valuable buried nuggets - Miners have no control over what’s there or how hard it is to separate it out - Data Mining seeks valuable information buried within massive amounts of data • Farmers cultivate to maximize yield 2 Data Mining vs. Data Farming • Miners seek valuable buried nuggets - Miners have no control over what’s there or how hard it is to separate it out - Data Mining seeks valuable information buried within massive amounts of data • Farmers cultivate to maximize yield - Farmers manipulate the environment to their advantage: pest control, irrigation, fertilizer, etc. 2 Data Mining vs. Data Farming • Miners seek valuable buried nuggets - Miners have no control over what’s there or how hard it is to separate it out - Data Mining seeks valuable information buried within massive amounts of data • Farmers cultivate to maximize yield - Farmers manipulate the environment to their advantage: pest control, irrigation, fertilizer, etc.
    [Show full text]